What is the role of genetic drift in phenotypic evolution, if any? Even if most evolutionary biologists have strong reasons to believe that genetic drift is a weak force in phenotypic evolution, in comparison to selection [69, 70], genetic drift can still influence trait variation that later becomes visible to selection . For example, the loss of a particular color morph might change the social selective environment, particularly when color is used in intra- or intersexual signaling [46, 71]. Genetic drift in sexually selected traits in isolated populations might lead to profound consequences upon secondary contact. Since sexually selected characters can diverge along a neutral line of equilibrium , genetic drift in sexually selected characters can result in sexual isolation between populations as a by-product of sexual selection within populations . An interaction between sexual selection and genetic drift is of particular interest in the present study as well as in other lizard species, since throat color in lizards has been found to be correlated with immunological condition and color could thus function as a sexually selected honest signal both within and between the sexes .
Comparisons of phenotypes from island and mainland populations played a central role in Ernst Mayr's theory of founder effect speciation [1, 74] and continues to inspire contemporary evolutionary biologists today . For instance, island biology studies of enigmatic color morphs in Dendrobates frogs have shown an impressive color morph diversity in the archipelago of Bocas del Toro, outside Panama, compared to mainland Central America . The discrete throat color phenotypes in P. gaigeae and the natural geographical replicates (i.e., different islet populations), as well as the variation in throat color throughout the Balkan clade of Podarcis where the sister species P. milensis does not exhibit throat color morphs whereas for example P. melisellensis has retained it, make this an excellent study system to address how selective processes operate in subdivided islet populations.
A role for phenotypic plasticity in determining the throat color morphs in P. gaigeae can not be excluded. If throat color morph would be entirely plastic with no heritable basis, the significant correlations between neutral genetic variation and the throat color morph frequencies for mainland-mainland and islet-mainland population pairs (Fig. 6A) would still need an explanation. One possibility might be that these correlations could reflect habitat differences between populations that have diverged to different degrees and at a rate that is proportional to molecular divergence. One could possibly argue for this possibility on the main island if there would be a clinal change in habitat. However, such a correlation would not necessarily be expected between the islet-mainland population pairs since the similarity in habitat between the islets and mainland locals is unlikely to be proportional to the time since the islets were isolated from the mainland. Rather, the environments on the islets depend more on local factors such as soil type and anthropogenic impact such as grazing pressure from goats (A. Runemark and E. Svensson, unpublished observations). Although we do not exclude the existence of genotype-by-environment interactions, a scenario with no heritable basis at all for the throat color polymorphism seems quite unlikely. With the caveat that we have no data from breeding experiments on the genetic basis of this throat color polymorphism, the discrete nature of these color morphs (Fig. 3), the strong correlation in throat color between the sexes (Fig. 4), our findings that none of the populations deviate from Hardy-Weinberg proportions with respect to the throat color morphs (Additional file 2), and the significant relationship between population divergence of the throat color morphs and neutral genetic divergence (Fig. 6A), is consistent with a heritable basis of throat color. Previous studies on several other lizard species have often found indications of a relatively simple genetic architecture of color morphs [30–34]. Although we do not know the exact number of loci involved in determining the throat color, the results are based on allele frequencies, and even if more than one locus would determine the throat color, it would not qualitatively change our conclusions, as long as there is some genetic basis to the polymorphism (e. g. a major-effect-gene). Based on the evidence above we tentatively assumed a simple genetic basis, and investigated if the data on the spatial distribution of these color morphs is consistent with the neutral expectation, and that stochastic factors alone might responsible for the population differentiation.
In the genus Podarcis several, if not most, species exhibit throat color polymorphism which indicates that these polymorphisms might have survived several lineage splitting and speciation events , (Fig. 1). When a polymorphism has been retained from a common ancestor it is highly unlikely to be entirely selectively neutral because strong overdominance or negative frequency-dependent selection is required to maintain all morphs in such systems [77–79], and selection on traits, including color polymorphisms, is usually a strong force . Thus, it is probable that there is long-term balancing selection that maintains the throat color morphs in P. gaigeae and other Podarcis-species. In addition, if throat color morph was selectively neutral it would be highly unlikely that all three alleles would be present on all mainland locations and two out of three islets. Consistent with such a scenario where selection acts in a stabilizing fashion, the frequencies of the throat color morphs were quite similar across all the mainland populations on Skyros (Fig. 2B). Interestingly, the orange throat color morphs were overrepresented on the islets compared to the mainland populations (Fig. 2B). These differences between mainland and islet populations could potentially be a result of different local selective environments on the islets or islet specific genotype by environment interactions if such exist, although we have no direct ecological data in support of this. Subdivided populations with limited effective population size are, however, also affected by genetic drift . Thus, the differences in throat color morph frequencies between the geographically close islet populations with similar habitats (Figs. 2A-B) could potentially partly be a result of stochastic processes.
Random fluctuations due to genetic drift can result in a form of environmental noise that can in turn fuel selection. For instance, when negative frequency-dependent selection operates in a system with environmental noise, perturbations in one direction can generate strong backlashes, which in turn can fuel selection and cause rapid evolutionary dynamics . When the selective signal is weak and environmental noise is high, stochastic factors might partly and temporarily overcome the selective signal (see for example ). Stochastic effects such as genetic drift have a larger impact when population size is small , as each stochastic death implies a larger change in the allele frequency distribution. Moreover, when one of the alleles is maintained at a low frequency at stable evolutionary equilibrium, selection may accelerate the loss of the rarest allele by random genetic drift . In several other color polymorphic systems, including species with continuous distributions and large population sizes, the frequency of the different color morphs sometimes shows patterns which are not distinguishable from the expectations under a scenario of genetic drift [7, 9, 23]. In addition to genetic drift, other forms of stochasticity can also operate and affect allele frequencies, even in large populations . These forms of stochasticity include demographic stochasticity and stochasticity generated by life-history variation among genotypes .
Three different lines of evidence suggest that genetic drift might partly be responsible for the divergence in throat color morph frequencies between the islets given that the throat color polymorphism is heritable (Fig. 2B). First, the effective population sizes on the islets are very low (Table 2; All Ne:s < 100). In such small populations, stochastic frequency fluctuations in selected alleles might under some periods temporarily override the effects of selection . With an effective population size of approximately 40 individuals, as is the case for one of the islets, there are only 80 throat color alleles present and random loss of only a few individuals could potentially have large effects on the overall population allele frequency. Second, the FST-values of traits subjected to directional selection are expected to differ significantly from the FST-values for neutral genetic variation [44, 45]. The FST for throat color morph frequency fell in the lower range of the distribution of the FST-values for neutral genetic variation (Fig. 6B). Comparisons of population divergence for different sets of loci are useful for identifying loci under directional selection , although the large confidence intervals which sometimes overlap zero can make it difficult to infer stabilizing selection with this approach . The overall high neutral FST-value for this lizard species (0.208) should, however, increase our statistical power to detect stabilizing selection in this system that might have favored a certain stable morph frequency equilibrium in all populations. Third, we found a significant positive correlation between the population differentiation in throat color and neutral genetic divergence (Fig. 6A). If selection would mainly explain the population morph frequency differences, we would not necessarily expect a change in genome-wide neutral genetic variation. Selection pressures to maintain a stable morph frequency across all populations might be the appropriate null expectation of population divergence for polymorphisms that transcend species boundaries, and when alleles are maintained by negative frequency-dependent selection [77–79]. If this is the case also for a trait like this color polymorphism, we would expect a pattern of population divergence for color morphs that would be significantly less than the neutral expectation at evolutionary equilibrium .
We do certainly not claim, based on these results, that genetic drift is the only evolutionary force operating on this color polymorphism. An additional role for selection and possibly for phenotypic plasticity can certainly not be excluded. Although indirect inferences about the action of selection are useful and have been used successfully by many workers in the past [15, 82–84], such indirect approaches suffer from several limitations, among them low statistical power [45, 81]. Failure to reject the drift-expectation for color morph divergence in this study (Fig. 6B) does thus not justify any strong claim that selection does not operate at all on these color morphs. Rather, we suggest that selection on these morphs, at least on the small islets, is not strong enough to result in a significant adaptive signal when we compared adaptive divergence and neutral divergence. Although genetic drift is usually considered by many population geneticists to be a weak force compared to selection [69, 70], we tentatively suggest that an interaction between genetic drift and local natural or sexual selection is most likely explanation for the spatial patterns in our data. Genetic drift could change trait frequencies when selection is temporally relaxed, an ecological scenario that is likely to be quite common and which might exist in this system, for instance, when lizard populations invade novel island environments with fewer predators. We can not disentangle the relative contributions of genetic drift and founder effects based on our findings in this study. Since the sea level in the Aegean Sea has been rising and sinking [52, 53] a scenario where lizards populations were isolated when land bridges were submerged and subsequently lost genetic variation due to low effective population sizes is probably more plausible than a pure founder event scenario, though.
The physical and biotic environments differ substantially among these small islets, which could generate novel selection pressures compared to the mainland. For instance, the predator faunas (birds and snakes) are more depauperate on the islets, and the seabird colonies on the islets might alter the nutrient cycles considerably, which in turn might affect the quantity and quality of the food sources of the lizards . It is possible that the islet populations have adapted locally to meet the requirements of these island specific environments, or that genotype-by-environment-interactions result in certain colour morphs becoming expressed more often in certain environments. The vegetation on the islands in the Skyros archipelago differs substantially even among nearby islets, which might partly reflect stochastic factors and reveal some degree of community drift . For instance, some common plant taxa are lacking entirely on some of the islets, whereas the abundance of other plant taxa is much higher on some of the islets than on the mainland . Although these different habitat features might potentially favor different throat color morphs on different islets, the mainland populations also differ in many environmental variables, such as type of substrate, grazing pressure, vegetation cover and predator faunas (A. Runemark & E. I. Svensson, unpublished observations).
After the geographic isolation of P. g. weigandi from P. g. gaigeae, the orange morph increased in frequency on Piperi (Figs. 2A-B), perhaps as a result of founder events, genetic drift, or because of local selection favoring that morph in this novel island environment. If the different throat color phenotypes are genetically correlated to other physiological, morphological or behavioral traits, which is the case in U. stansburiana [26, 28, 32, 85] or in immune response as is the case for P. muralis , one or several of the throat color morphs in the ancestral population could initially and due to chance alone, have been better adapted to the novel environments. Such "pre-adapted" morphs might then subsequently rise in frequency through phenotype sorting [86, 87], either by selection alone or due to an interaction between selection and genetic drift. The loss of one or more color morphs has occurred several times in different populations of U. stansburiana [46, 71], and it might also operate in the Balkan clade of Podarcis (see e. g. P. milensis in Fig. 1; ).
If genetic drift can affect phenotypic divergence in characters such as throat color which are often genetically correlated with other traits (see discussion above), genetic drift could potentially indirectly also change the optimal strategies for the remaining morphs since the selective environment will change if one morph and its strategy is removed. For instance, populations of U. stansburiana, which have lost morphs and become monomorphic, differ from the polymorphic populations with respect to sexual size dimorphism and potentially sexual selection . Thus, the loss of one or several morphs might cause rapid divergence following a change in social selective environment in lizards. Polymorphic types within a population can thus potentially serve as starting material for new species [47, 88, 89] and the potential loss of the yellow throat color morph in the subspecies P. g. weigandi is interesting from this point of view.